gsa-displacy-demo

Setup

• Anaconda for package handling
• models compiled with Linux
• recommend Use a compute EC2 like c5xlarge
• pip installed packages in the requirements-full.txt
• git-lfs for hugging face models

Make sure to have conda installed

• create new conda environment: conda create --name=gsa_displacy_demo python=3.9
• activate new conda env: conda activate gsa_displacy_demo
• enable shell script to be executed: chmod +x setup.sh
• install dependencies using shell script: bash setup.sh

Alternative Setup

Use this when you are unable to python -m install the spacy models

• create new conda env: conda create --name=gsa_displacy_demo python=3.9
• activate new conda env: conda activate gsa_displacy_demo
• pip install requirements: pip install -r requirements.txt
• manually download spacy models: Download the following models from link
  • en_core_web_sm
  • en_core_web_lg
  • en_core_web_trf
• manually download GSA in-house trained models:
  • Email classifier from digital cx analysis GovCXAnalyzer/ditgitalcx
  • News relevance classifier from Strategic Atlas pilot
• move model tar files to project subfolder called 'models/', and extract each tar

Run

to run:

• activate conda environment (if note already activated): conda activate gsa_displacy_demo
• start streamlit app: streamlit run main.py

Contact Info

• Email isabel.metzger@gsa.gov (Data Scientist, GSA OCFO Analytics and Decision Support Team) with any questions.